Upload
vinicius-cabral
View
218
Download
0
Embed Size (px)
Citation preview
7/28/2019 Paper Wesley Pacheco
1/23
Horizontal Stratifcation o the Soil in Multi-Layer
Using Non-Linear Optimization
Estratifcao Horizontal do Solo em Vrias Camadas
Usando Otimizao No-Linear
Wesley Pacheco Calixto
Electromagnetism and Electric Grounding Systems Nucleus Research and
Development - Federal University o Uberlandia - Department Electrical
Engineering, Av. Joao Naves de Avila 2160 Campus Santa Mnica Bloco 1E - Sala
1E20 CEP 38400-902 Uberlandia, Minas Gerais, Brazil
Luciano Martins Neto
Electromagnetism and Electric Grounding Systems Nucleus Research and
Development - Federal University o Uberlandia - Department Electrical
Engineering, Av. Joao Naves de Avila 2160 Campus Santa Mnica Bloco 1E - Sala
1E20 CEP 38400-902 Uberlandia, Minas Gerais, Brazil
Marcel Wu
Electromagnetism and Electric Grounding Systems Nucleus Research and
Development - Federal University o Uberlandia - Department Electrical
Engineering, Av. Joao Naves de Avila 2160 Campus Santa Mnica Bloco 1E - Sala
1E20 CEP 38400-902 Uberlandia, Minas Gerais, Brazil
Hiplito Barbosa Machado Filho
Electromagnetism and Electric Grounding Systems Nucleus Research and
Development - Federal University o Uberlandia - Department ElectricalEngineering, Av. Joao Naves de Avila 2160 Campus Santa Mnica Bloco 1E - Sala
1E20 CEP 38400-902 Uberlandia, Minas Gerais, Brazil
Resumo: O propsito principal deste trabalho apresentar uma metodologia
que utiliza um processo de otimizao no-linear na obteno dos parmetros
da estraticao horizontal do solo em vrias camadas. O mtodo utiliza uma
7/28/2019 Paper Wesley Pacheco
2/23
68
Revista Cincias Exatas e Naturais, Vol.11 n 1, Jan/Jun 2009
curva de resistividade experimental proveniente de leituras eitas no solo. Esta
curva experimental comparada com uma outra curva de resistividade terica
produzida pelo processo de otimizao. A curva terica embasada no algoritmo
de Sunde que exatamente o processo inverso da estraticao horizontal do solo
em vrias camadas. De posse das duas curvas, pode-se mensurar o erro produzido
na estraticao horizontal do solo. Os parmetros a serem estimados so os
nmeros de camadas N , a resistividadei
e a espessura ih de cada camada.
Palavras-chave: Resistividade Eltrica do Solo; Estraticao do Solo; Otimizao
No-Linear.
Abstract: The main purpose o this paper is to present a methodology to obtain
the stratication o the soil into horizontal multi layers rom measuring results,
obtained rom the Wenners method. A mathematical modeling is presented
combining the horizontal stratication o the soil method in several layers, with
the non-linear optimization o Quasi-Newton method. The process initiates,
calculating the resistivity o the rst layer in a non-traditional way, and then
calculating the resistivity and the thickness o the ollowing layers, based onthe Sundes algorithm, which is exactly the inverse process o soil stratication
into multi layers. The parameters to be optimized in this methodology are: the
resistivityi
, the thickness ih and the number o layers N . The results achieved
point out advantages in comparison to other known methods.
Key words: Electrical Resistivity o Soil; Stratication o Soil; Non-Linear
Optimization.
1 Introduction
In an electric grounding project, it is essential to know the soil behavior,
regarding to its electrical conduction properties. In metals, this property is
represented by the value o its electric resistivity, which varies according to the
7/28/2019 Paper Wesley Pacheco
3/23
CALIXTO, W. P. et al.
69
metal temperature, impurities etc. However, variations on the electrical resistivity
o electric conductors are relatively small, maintaining a great homogeneity
along its dimensions. In case o the soil, the electrical resistivity variations are
not only extremely large, as well as it has a heterogeneous behavior along its
dimensions [1].
This characteristic represents a major obstacle or designers o electric
grounding. Due to this diculty, a model that represents the soil by its
electricity conduction behavior is developed. The most common model is the
one which represents the soil as horizontal layers, where each one o them has
the same electrical resistivity throughout its points. Since an electric groundingextension does not overtake more than a hundreds o meters, this model can
be applied with a good accuracy in the major part o the electric grounding
designs, justiying its wide use.
With a soil stratication model adopted, it is only necessary to dene the
area to be stratied i.e. to a certain area addressed to build an electric grounding,
it is necessary to know, (i) with how many layers its soil can be represented, and
(ii) the values o thickness and resistivity o each soil layers.
Frank A. Wenner [2], carried out a measuring method, which the resultscontains the answers o those prior questions. Although this measuring method
has some fuctuations, one o them will t best in the electric grounding project
[3]. The method conguration is shown in g. 1, where our align and equally
spaced electrodes are inserted into the soil.
Figure 1. Wenners Method
Source: The authors
7/28/2019 Paper Wesley Pacheco
4/23
70
Revista Cincias Exatas e Naturais, Vol.11 n 1, Jan/Jun 2009
When a current I is orced into the external electrodes, the voltage V can
be measured through the internal electrodes [4]. From [5, 6], the resistivity o a
homogeneous soil is given by (1).
( )
( ) ( ) ( )
a
aRa
a a
a P a P
2 2 22
4=
2 21+ -
+ 2 2 + 2
(1)
Where a is the spacing between electrodes, considering the Wenners
method, P is the depth in which the electrodes are immersed into the soil and R is
the measured soil resistance. I the soil is homogeneous, when the distance a varies,
the relationI
VR = becomes inversely proportional, related to a , so remains a
same value, which characterizes a homogeneous soil with resistivity .
However, in a heterogeneous soil, when a varies, the relationI
VR = can
also varies, but not in the same way as in the prior case. Thereore, the resistivity
has dierent values or each distance a . This variable resistivity is going to becalled apparent resistivity o the soil a , which is a unction o a , i.e., )(aa .In g. 1, varying the distance a , with a xed reerence point, it is possible to rise
the experimental curve )(aa .This curve stores inormation about the number olayers with their respective electrical resistivity and thickness. The main purpose
o this paper is to show an ecient and accurate alternative way to stratiy the
soil, i.e., determinate the number o layers and its respective values o electrical
resistivity and thickness, using the Wenners experimental curve, )(aa . Themethodology employed applies a non-linear optimization process based on the
mathematical modeling deduced rom the Sundes algorithm [7].
As it is presented throughout the paper, advantages are ound in using this
method o soil stratication rather than other methods already known.
2 Soil Stratifcation
The vast majority o electric grounding projects are a combination o vertical
and horizontal ground conductors. Knowing the soils layers, with their respective
7/28/2019 Paper Wesley Pacheco
5/23
CALIXTO, W. P. et al.
71
resistivities and thicknesses, the depth that the vertical conductor should achieve
can be dened, assuring the best use o the soil.
I there is a thick rst layer with low resistivity in comparison to the others,
it is probably that the grounding should be built supercial, being more indicated
the use o horizontal conductors [8]. I the opposite occurs, vertical ground
conductors should be employed, leaving the horizontal conductors only with the
role o electric linking. Whether are known the soil structure values, it is possible
to model the grounding geometry, i.e., to work with vertical and horizontal
conductors in a most ecient way, so much regard with the grounding resistance,
as well as in the potentials, mainly the supercial ones.
When interpreted along with the values o soils layers, those supercial
potentials can dene certain projects criteria. A common case is the vertical
conductors. I it belongs to thick and low resistive layer, the current that fows
rom it tends to spread itsel throughout the rst layer, producing surace potentials
until great distances [9, 10]. These phenomena smoothes the potentials curves,
transorming the vertical conductor application less dangerous when related to
step voltages[11].
However, i this same supercial layer has high resistivity values, thecurrent will fow deeper, producing supercial potentials that are reduced in small
distances [12], causing a abruptly decreasing in the potentials curves [13]. In this
case, the vertical conductor employ, can become more dangerous regarding to
step voltages.
Another uncountable number o analyses can be done relating the
grounding conductors with the soils structure in horizontal layers [14]. These
analyses dene extremely useul orientations criteria to the development o a good
electric grounding project.
3 Mathematical Modeling
The soil stratication proposed on this paper has on its basis, the idea o
comparison between the experimental Wenners curve ( )aE a and the theoretical
7/28/2019 Paper Wesley Pacheco
6/23
72
Revista Cincias Exatas e Naturais, Vol.11 n 1, Jan/Jun 2009
Wenners curve ( )aT a . This last one, created rom a known soil structure,i.e., number o layers N , thickness ih and resistivity i , o a generic layer i .
Thereore, the basic idea is: when one theoretical curve equal to the experimental
curve is obtained, considering a maximum admissible error, the values o the
soil structure that generates this theoretical curve correspond to the desired soil
stratication. From [7], (2) can be written.
( ) ( ) ( ) ( )[ ]a na a N m J ma J ma
1 0 00= 2 2 (2)
The rst layer resistivity is 1 , m is the auxiliary integration variable, 0J
the Bessel unction o type one and order zero, and nN is the Sundes characteristic
unction. I the soil structure is known, it is possible to obtain its characteristic
unction by Sundes algorithm [7], i.e., N ,i
and ih . In Sundes algorithm,
equality (3) can be adopted.
E=e-2mh1 (3)
In (3) it is presented a variable change rom m to E, which enables
to express the characteristic unction nN in E, and, in this particular case, a
polynomial expression described in (4) was chosen.
nN = A + A E + A E +...2
0 1 2
(4)
It is possible to veriy by Sundes algorithm that or 0=E it is obtained
1=nN and, thereore 1=0A . So (4) becomes (5).
N =1+ A E + A E +...2
n 1 2 (5)
Assuming, at rst, a certain number o terms j to (5), we can choose
j values or E, remembering that 1
7/28/2019 Paper Wesley Pacheco
7/23
CALIXTO, W. P. et al.
73
It is possible by an iterative process to obtain the best values o j , minimizing
numerical errors generated by the polynomial decomposition. I the values o kA are
known, the equation (5) express nN as unction oE. Considering this data, it
is possible to solve the integral in (2) analytically. Solving the integral in (2) and
using some algebraic manipulation (6) is obtained.
( ) ( )j
a 1 k k
k=1
r a = r . + A .S a 1 2 (6)
Where:
( )kS a = S +S 1 2 (7)
In which:
S =.b .k +
1 2 2
1
4 1(8)
S = . b .k +2 2 21
2 1(9)
a
h1= (10)
To obtain the theoretical curve ( )aT a equation (6) is used. To initiate theoptimization process, saving processing time, it is interesting to obtain the initial
values o the process coherent with the experimental Wenners curve ( )aE a . Thisbecame possible determining at rst, the electric resistivity o the rst layer. For this,
the unction )(aa o (6) may be written assuming pairs o values o a , i.e., 1a and
2a .
( ) ( )j
a 1 k k
k
a = A .S a 1 1=1
. 1+ 2 (11)
7/28/2019 Paper Wesley Pacheco
8/23
74
Revista Cincias Exatas e Naturais, Vol.11 n 1, Jan/Jun 2009
( ) ( )j
a k k
k
a = A .S a 2 1 2=1
. 1+ 2 (12)
Dividing and subtracting (11) and (12), expressions (13) and (14), are
respectively obtained.
( ) ( )( )( )
( )
j
k ka k
ja
ak k
k
A S aar a
a A S a
1
1 =1
1,2
22
=1
1+2 .= =
1+ 2 .
(13)
( ) ( ) ( ) ( ) ( )[ ]j
a a a k k k
k
d a a a A S a S a 1,2 1 2 1 1 2=1
= - = 2. . - (14)
From (13) and (14), equations (15) and (16) can be obtained, respectively as:
j
k k a
k
A M a r a 1,2 1,2=1
. ( ) = ( )-1 (15)
( ) ( )j
a
k k
k
d aB P a
1,21,2=1
. =2
(16)
Where:
( ) ( ) ( ) ( )k k a k M a S a r a S a1,2 1 1,2 2= - . (17)
k kB A1= . (18)
( ) ( ) ( )k k kP a S a S a 1,2 1 2= - (19)
Thus,1 can be obtained rom (15) and (16). This process is explained in
the next section.
7/28/2019 Paper Wesley Pacheco
9/23
CALIXTO, W. P. et al.
75
4 Methodology
When 1a and 2a are both modied by the algorithm iteration, expressions
(15) and (16) represents two system o equations, where the unknown variables
are kA , kB and 1h .
Firstly, the goal is to get only the rst layer resistivity 1 . This layer can
be imagined as composed by two virtual horizontal layers, both with resistivity
1 . Considering that in the real world, the thickness o the rst layer is usually
superior to 10 cm, one can adopt the thickness o the rst virtual layer as 10 cm.
Thereore, or purposes o 1 calculation, 1h can be assumed to 10 cm, and thus,rom (15) and (16), 1 is obtained.
Knowing the1 value, the characteristic unction nN can be obtained
by Sundes algorithm. So, the kA coecients also can be obtained by using the
polynomial (5). From expression (6) the theoretical Wenners curve ( )aT a canbe draw. In order to generate this curve it is necessary to adopt values o N , i
and ih , remembering that 1 was denitely obtained. The comparison between
the experimental Wenners curve and the theoretical Wenners curve shows i the
assumed values oN, i and ih represent a good stratication.At this point, a non-linear optimization method is employed, so that by
an iterative process, the set o values N,i
and ih are being modied, until a
theoretical curve as close as possible o the experimental curve is obtained.
The usual employed methodologies estimate the number o layer N,
regarding the number o infections p presented at the experimental curve o
apparent resistivity o the soil i.e. 1= +pN . At the proposed methodology, thenumber N o layers is a parameter to be optimized.
The optimization method adopted is called Quasi-Newton [15], already
known as the Secants Method its use is justied once it brings the simplicity o
the Gradient method as much as the quickness o the the Newtons method [16].
Rather than calculating the inverse o Hessian matrix, as it is done in the Newtons
method, this inverse process is approached by nite iterations using the rst order
dierentiations, as it is done in the gradient method [15].
7/28/2019 Paper Wesley Pacheco
10/23
76
Revista Cincias Exatas e Naturais, Vol.11 n 1, Jan/Jun 2009
Assuming that aEx is the experimental resistivity curve and aTh the
theoretical resistivity curve, the evaluation unction )(xf which is a metric
is dened by:
( ) aEx aThaEx
x
-
= (20)
The Quasi-Newton algorithm updates the matrix with the parameters N ,
i and ih to be optimized, minimizing the deviation value )(xf ound. The N
value, is a integer number, distinct rom the other values oi
and ih , due to this,
there is a dierent treatment or the N value optimization.
The experimental and theoretical curves were obtained rom the interpolation
o the soil apparent resistivity values. This interpolation is made with groups o
ew points, obtained through a lower degree polinomial, imposing conditions or
the approach unction to be continuous and having continuous derivatives until
a certain order [17]. This makes the curve to not have peaks and neither abruptly
changes o curvature in its knots, sotening the curve.
5 Results
This section presents the results rom six dierent soil stratications process.
Among them, two has resulted in a two layers stratication, other one has resulted
in a three layers, and the rest o them resulted in a our layers stratication. Small
deviation values are obtained between the experimental and theoretical curves o
apparent resistivity o the soil. The Wenners curves, which are illustrated rom 2 to
7 were obtained through an interpolation o the experimental curves points. The
resistivity experimental curve, proceeds rom eld measuring, and the theoretical
curve, rom the Sundes algorithm.
In the Wenners method application, several measures are done in the
eld at dierent directions and, in general, some are discarded ater analysis.
As the measure harvested in the eld refects the soil real conguration and,
what we aim here, is to demonstrate a new methodology to stratiy the soil, in
7/28/2019 Paper Wesley Pacheco
11/23
CALIXTO, W. P. et al.
77
this work, the measures were harvested just in one direction, or the Wenners
method application.
A larger amount o measured points, i.e. or a larger amount o a values
utilized to produce the experimental curve o apparent resistivity o the soil, heads
to lowest deviations between the experimental and theoretical curves o apparent
resistivity o the soil.
For the process o Soil Stratication 1, the data o the experimental and
theoretical curves o apparent resistivity o the soil, are given at tab. 1, with the
Wenners curve represented in g. 2 and the results o the stratication process
(thickness and resistivity o each layer), given at tab. 2. The same way is used or
the others stratications processes, being tab. 3 and tab. 4 with the g. 3, related
to the Soil Stratication 2, the tab. 5 and tab. 6 with the g. 4 to Soil Stratication
3, tab. 7 and tab. 8 with the g. 5 to Soil Stratication 4.
For stratications numbers 5 and 6, two distinct soil stratication methods
were used. One o them is the proposed method, and the other is known as the
complex image method [18, 19].
At tab. 9 are listed the theoretical and experimental values o apparent
resistivity o the soil rom the the stratication process number 5, using the
proposed method. tab. 10 contains the theoretical and experimental values o
apparent resistivity o the soil rom the process o stratication number 5, using
the method o the complex image. g. 6 presents the curves interpolated with the
points listed at tab. 9 and tab. 10, the results o thickness and resistivity o each
layer, ound by the proposed method and the method o the complex image, are
presented in tab. 11.
The same sequence described above were made or the soil stratication
process number 6. The theoretical and experimental values o soil resistivity
ound by the proposed method are listed at tab. 12, and the values ound with the
complex image method are listed at tab. 13, the interpolated curves are plotted on
g. 7, and the results o this stratication, ound with the proposed method and
the complex image method, are presented at tab. 14.
7/28/2019 Paper Wesley Pacheco
12/23
78
Revista Cincias Exatas e Naturais, Vol.11 n 1, Jan/Jun 2009
5.1 Soil Stratifcation 1
Table 1. Stratifcation 1 - Experimental and Theoretical Curves
a Experimental Theoretical Deviation
(m ) Curve ( m ) Curve ( m ) (% )
1.0 641.83 653.43 1.81
2.0 996.62 1014.29 1.77
4.0 1437.62 1456.71 1.33
8.0 1887.08 1850.27 1.95
16.0 2091.32 2098.57 0.35
Figure 2: Curve o Apparent Resistivity o the Soil rom Stratifcation 1
Table 2. Stratifcation 1 - ResultsLayer Layer Layer
NumberThickness (
m) Resistivity (
m)
st1 0.72 408.99 nd2 inf 2257.95
7/28/2019 Paper Wesley Pacheco
13/23
CALIXTO, W. P. et al.
79
Through the tab. 1, one can notice that the major deviation between
the theoretical and the experimental curve is 1.95% which corresponds topoint 8 meters .
5.2 Soil Stratifcation 2
Table 3. Stratifcation 2 - Experimental and Theoretical Curves
a Experimental Theoretical Deviation
(m ) Curve ( m ) Curve ( m ) (% )
1.0 3582.92 3663.81 2.26 2.0 3354.57 3258.08 2.88 4.0 1872.05 1927.73 2.97 8.0 2104.58 2106.74 0.10
16.0 2322.47 2210.28 4.83
Figure 3. Curve o Apparent Resistivity o the Soil rom Stratifcation 2
7/28/2019 Paper Wesley Pacheco
14/23
80
Revista Cincias Exatas e Naturais, Vol.11 n 1, Jan/Jun 2009
Table 4. Stratifcation 2 - Results
Layer Layer Layer
Number Thickness (m ) Resistivity ( m )st1 1.31 4120.35nd2 2.21 1089.34rd3 2.75 4220.83th4 inf 2007.31
Through tab. 3, one can notice that the major deviation between the theoretical
and experimental curve is 4.83% which corresponds to point 16 meters.
5.3 Soil Stratifcation 3
Table 5. Stratifcation 3 - Experimental and Theoretical Curves
a Experimental Theoretical Deviation
(m ) Curve ( m ) Curve ( m ) (% )
1.0 16841.29 16120.21 4.28 2.0 20715.20 19966.26 3.61 4.0
15381.67
15422.58
0.27
8.0 9483.99 8999.18 5.11 16.0 24970.15 23716.15 5.02
Figure 4. Curve o Apparent Resistivity o the Soil rom Stratifcation 3
7/28/2019 Paper Wesley Pacheco
15/23
CALIXTO, W. P. et al.
81
Table 6. Stratifcation 3 - Results
Layer Layer LayerNumber Thickness (m ) Resistivity ( m )
st1 1.06 14315.09 nd2 0.78 21814.81rd3 inf 13535.38
Through tab. 5, one can notice that the major deviation between the theoretical
and experimental curve is 5.11% which corresponds to point 8 meters .
5.4 Soil Stratifcation 4
Table 7. Stratifcation 4 - Experimental and Theoretical Curves
a Experimental Theoretical Deviation
(m ) Curve ( m ) Curve ( m ) (% )
1.0 9560.68 9281.51 2.92 2.0 11135.90 11105.60 0.26 4.0 7328.86 7073.66 3.48
8.0 11591.73 11589.17 0.02 16.0 11990.35 11994.37 0.03
Figure 5. Curve o Apparent Resistivity o the Soil rom Stratifcation 4
7/28/2019 Paper Wesley Pacheco
16/23
82
Revista Cincias Exatas e Naturais, Vol.11 n 1, Jan/Jun 2009
Table 8. Stratifcation 4 - Results
Layer Layer LayerNumber Thickness (m ) Resistivity ( m )
st1 0.59 8126.57 nd2 3.30 9174.68 rd3 4.58 17388.55 th4 inf 10257.32
Through tab. 7, one can notice that the major deviation between the theoretical
and experimental curve is 3.48% which corresponds to point 4 meters .
5.5 Soil Stratifcation 5
Table 9. Stratifcation 5 - Experimental and Theoretical Curves (Quasi-Newton
Method)
a Experimental Theoretical Curve Deviation
(m ) Curve Quasi-Newton (% )
( m ) Method ( m )
1.0 1338.76 1324.15 1.09
2.0 1360.86 1438.28 5.69
4.0 1472.59 1459.73 0.87
8.0 1300.99 1301.90 0.07
16.0 1078.87 1075.90 0.28
Table 10. Stratifcation 5 - Experimental and Theoretical Curves (Complex image
Method)
a ExperimentalTheoretical
CurveDeviation
(m ) Curve Complex Image (% )
( m ) Method ( m )
1.0 1338.76 1373.34 2.58
2.0 1360.86 1379.78 1.39
4.0 1472.59 1389.64 5.63
8.0 1300.99 1302.39 0.11
16.0 1078.87 1073.92 0.46
7/28/2019 Paper Wesley Pacheco
17/23
CALIXTO, W. P. et al.
83
The values o each layer resistivities1
, 2 , 3 and 4 , and the respectivesthickness values 1h , 2h , 3h and 4h obtained rom each stratication method, arecompared at tab. 11.
Figure 6. Curve o Apparent Resistivity o the Soil rom Stratifcation 5
Table 11. Stratifcation 5 - Results
LayerQuasi-Newton Complex Image
Method Method
1 1137.94 1371.99
2 1538.89 1866.52 3 336.23 137.99 4 9208.28 962.96 1h 0.52 3.74
2h 7.78 7.11 3h 7.31 7.66
4h inf inf
7/28/2019 Paper Wesley Pacheco
18/23
84
Revista Cincias Exatas e Naturais, Vol.11 n 1, Jan/Jun 2009
As can be observed at the tab. 9, tab. 10 and g. 6, there is a great proximity
between the results obtained by the stratications made with the proposed method
and the complex image method. However, smaller deviation between the curve
o apparent resistivity o the soil obtained with the proposed method and the
experimental one, were ound.
5.6 Soil Stratifcation 6
Table 12. Stratifcation 6 - Experimental and Theoretical Curves (Quasi-Newton
Method)
a Experimental Theoretical Curve Deviation
(m ) Curve Quasi-Newton (% )( m ) Method ( m )
2.0 3389.00 3389.17 0.01 4.0 1900.00 1609.22 15.30 8.0 585.00 585.05 0.01
16.0 568.00 574.55 1.15 32.0
823.00
771.82
6.22
Table 13. Stratifcation 6 - Experimental and Theoretical Curves (Complex Image
Method)
a Experimental Theoretical Curve Deviation
(m ) Curve Complex Image (% )
( m ) Method ( m )
2.0 3389.00 3226.46 4.80
4.0 1900.00 2294.48 20.76
8.0 585.00 1105.38 88.95
16.0 568.00 690.07 21.49
32.0 823.00 640.72 22.15
7/28/2019 Paper Wesley Pacheco
19/23
CALIXTO, W. P. et al.
85
Figure 7. Curve o Apparent Resistivity o the Soil rom Stratifcation 6
Table 14. Stratifcation 6 - Results
LayerQuasi-Newton Complex Image
Method Method
1 99905.15 3550.00
2 570.77 630.00 1h 0.36 3.10
2 inf inf
The values o each layer resistivities1 and 2 , and the respectives
thickness values 1h and 2h obtained rom each stratication method, are
compared at tab. 14.
As can be observed at the tab. 12, tab. 13 and g. 7, there is a great proximity
between the results obtained by the stratications made with the proposed method
and the complex image method. However, smaller deviations between the curve o
apparent resistivity o the soil obtained experimentally and through the proposed
method, were ound, showing a improvement in the proposed stratication
method, when compared with the complex image method, which is a very usual
employed method, in soil stratications processes.
In the soil stratication process number 6, the number o layers N was
orced into 2=N . Analyzing the curve o experimental apparent resistivity o the
7/28/2019 Paper Wesley Pacheco
20/23
86
Revista Cincias Exatas e Naturais, Vol.11 n 1, Jan/Jun 2009
soil, g. 7, it can be notice that the number o curve infections 2=p , using the
described equation in section 4, 1= +pN , would be already enough to estimate anumber o layers 2>N . At tab. 12 and tab. 13 the listed deviations presented to be
larger than the acceptable in grounding projects. This horizontal soil stratication
is a typical case o stratication, where the number o layers N is orced in a xed
number, in this case, 2=N . Whether this stratication was made considering a
number o 3=N layers, the deviations would have a considerably decrease, both
to the two applied methods. This act, evidences the necessity o a methodology that
optimizes the number o layers N in the soil stratications.
Nevertheless, even orcing the number o layers to 2=N , it is observed
through tab. 12, tab. 13 and g. 7 that the proposed method achieved lower
deviations values than that obtained through the complex image method.
6 Conclusion
This paper has presented a methodology and a mathematical modeling or
the soil stratication in multi layers, where the number o layers, the resistivity
and the thickness o each layer are calculated and optimized rom the experimental
curve o apparent resistivity o the soil.
Several cases were processed. In all cases in which the stratication was
made by the proposed methodology, the results have presented lower or equal
deviations to those ound in the pertinent literature. This is explained by the acts
that diers this methodology, i.e.:
(i) The most common method employed to determinate the rst layer
resistivity is the extrapolation o Wenners experimental curve, where )(aa isextrapolated until the point where a is equal to zero. Knowing that a exploitationmethod is purely mathematical, when the curve )(aa does not have a goodbehavior, something that requently occurs, the obtained results can reach errors
which overtakes 200% . A such order error is unacceptable or a1 value,
invalidating the rest o the stratication. Some error at1 can be tolerable, but,
7/28/2019 Paper Wesley Pacheco
21/23
CALIXTO, W. P. et al.
87
depending on its order, it can propagate throughout the stratication, turning
its accuracy not reliable. In the proposed methodology,1 is obtained rom the
derivations expressions o (6), representing the physical signication in attendance
with Wenners measuring method. This ensures that1 accuracy depends mostly
on the adopted model to represent the soil.
(ii) Soil stratications considering a xed number o layers are common.
Actually, soil stratication involves three unknown variables: number o layers,
resistivity and thickness o each layer. Thereore, the number o layers is not a given
date o the problem, but an unknown variable. By the proposed methodology, the
set o unknown variables are: N ,i
and ih .
Finally, the accuracy obtained at this work, opens up to new investigations
possibilities. I the stratication accuracy is tied up to the project and the electric
grounding construction, there is no sense in using errors lower than one unit
percentage. An error around 10% is perectly easible, constructions imperections
o electric grounding justiy this order o magnitude. Still, there is something
more to be analyzed. In electrical engineering, the stratication o the soil or
electric grounding construction project, adopts the soil as a model, considering
its electrical conductibility in horizontally and uniormly layers. Although, some
soils behave dierently. This explains why sometimes the accumulated error o
stratication is not due only to the numeric process o curve adjustment. One
parcel o this error, sometimes the major one, can be caused by the act that the
soil does not have uniormity in its horizontal layers.
Restrictions on the proposed method can appear with the existence o
a vertical or oblique rit in the layer, this is enough to interere on the results
o Wenners method. In this situation, a very accurate stratication process is
important. I it is assured that the error made at the curve adjustment is small, but a
high error is still veried in the values obtained at the stratication, it is concluded
that the soil model made by horizontal layers is not applicable to this type o soil.
So, the great amount o small errors that appear in this paper, motivates urther
studies in order to obtain lowest errors on soil stratication, what justies, or
example, a more appropriated study on the optimization process.
7/28/2019 Paper Wesley Pacheco
22/23
88
Revista Cincias Exatas e Naturais, Vol.11 n 1, Jan/Jun 2009
Acknowledgment
This article results rom the researches o a P&D project (Development o a
Philosophy o Electric Grounding Project with Practical Extension) sponsored by
Energetic Company o Goias - CELG jointly with The Brazilian National Agency
o Electric Power - ANEEL.
Reerences
[1] IEEE Std 81-1983, ``Guide or Measuring Earth Resistivity, groundImpedance, and Earth Surace Potentials o a Ground Systems. Institute o
Electrical and Electronics Engineers, Inc. Revision o IEEE Std 81-1962.
[2] F. A. WENNER, ``Method o Measuring Earth Resistivity. Bulletin o the
National Bureau o Standards, Washington D.C., v. 12, 1916.
[3] H. YANG; J. YUAN; W. ZONG, ``Determination o Three-Layer Earth
Model rom Wenner Four-Probe Test Data. IEEE T. Magn., v. 37 n 5 p. 3684-
3687 2001.
[4] P. J. LAGACE; M. H. VUONG; M. LEFEBVRE; J. FORTIN, ``Multilayer
resistivity interpretation and error estimation using electrostatic images. IEEE T.
Power Deliver, v. 21, n. 4, p. 1954-1960, 2006.
[5] IEEE Std 142-1991, `Recommended Practice or Grounding o Industrial
and Commercial Power Systems. Institute o Electrical and Electronics
Engineers, Inc.
[6] IEEE Std 1100-1992, ``Recommended Practice or Powering and
Grounding Sensitive Electronic Equipment. Institute o Electrical and
Electronics Engineers, Inc.
[7] E. D. SUNDE, ``Earth conduction eects in transmission systems.
MacMillan & Company, London, 1968.
7/28/2019 Paper Wesley Pacheco
23/23
CALIXTO, W. P. et al.
89
[8] E. ORELLANA, ``Prospeccion Geoelectrica en Corriente Continua.
Paranino, Madrid, 1974.
[9] Y. L. CHOW; J. J. YANG; K. D. SRIVASTAVA, ``Grounding Resistance o
Buried Electrodes in Multi-Layer Earth Predicted by Simple Voltage Measurements
along Earth Surace: A Theoretical Discussion. IEEE T. Power Deliver, v. 10
1995.
[10] F. P. DAWALIBI; D. MUKHEDKAR, ``Ground Electrode Resistance
Measurements in Nonuniorm Soils. IEEE T. Power Ap. and Syst., v. PAS-93,
1974.
[11] F. P. DAWALIBI et. al., ``Optimum Design o Substation Grounding in a
Two Layer Earth Structure. IEEE T. Power Ap. Syst., Part I-II-III vol. PAS-94
1994.
[12] T. TAKAHASHI, `Calculation o Earth Resistance or a Deep-Driven Rod in a
Multi-Layer Earth Structure. IEEE T. Power Deliver, v. 7, 1991.
[13] A. L. KINYON, ``Correlation o Measured and Calculated Substation
Ground Grid Resistance. AIEE transactions, paper 59-806, nov 1959.
[14] R. W. RYDER, ``Earth Principles and Practice. Pitman and Sons,London, 1969.
[15] J. E. DENNIS; R. B. SCHNABEL, `Numerical Methods or Unconstrained
Optimization and Nonlinear Equations. SIAM, 1996, 378 p.
[16] M. S. BAZARAA; H. D. SHERALI; C. M. SHETTY, ``Nonlinear
Programming: Theory and Algorithms. Wiley, 1966, 853 p.
[17] E. ISSACSON; H. B. KELLER, ``Analysis o Numerical Methods. John
Wiley and Sons, 1993.
[18] Y. L. CHOW; J. J. YANG; K. D. SRIVASTAVA, ``Complex Images o
a Ground Electrode in Layered Soil. J. Appl. Phys., v. 71, p. 569-574, 1992.
[19] B. ZHANG; X. CUI; L. LI; J. HE, ``Parameter Estimation o Horizontal
Multilayer Earth by Complex Image Method. IEEE T. Power Deliver, v. 20 n.
2, p. 1394-1401, 2005.